zouroboros-introspect

name	zouroboros-introspect
description	Self-diagnostic health scorecard for the Zouroboros system. Measures memory recall quality, knowledge graph connectivity, swarm routing accuracy, eval calibration, procedure freshness, and autoloop efficiency. Outputs a ranked scorecard with improvement opportunities. Run daily or on-demand to feed the self-enhancement pipeline.
compatibility	Created for Zo Computer
metadata	{"author":"marlandoj.zo.computer","version":"1.0.0","phase":"Zouroboros Phase 1 — Introspection"}

Zouroboros Introspect

Self-diagnostic agent that measures health across all Zouroboros subsystems and produces a ranked scorecard of improvement opportunities.

Usage

bun Skills/zouroboros-introspect/scripts/introspect.ts [--json] [--store] [--verbose]

Flags

Flag	Description
--json	Output scorecard as JSON instead of formatted text
--store	Store the scorecard as a memory episode with outcome and entity tags
--verbose	Show detailed per-check output

What It Measures

Metric	Source	Target	Weight
Memory Recall	eval-continuation.ts	≥ 85% pass rate	0.25
Graph Connectivity	graph.ts knowledge-gaps	≥ 80% linked facts	0.15
Routing Accuracy	Episode analysis (predicted vs actual)	≥ 85%	0.20
Eval Calibration	Stage 3 override rate from eval reports	≤ 15% override	0.15
Procedure Freshness	Procedures not evolved in 14+ days	≤ 30% stale	0.15
Episode Velocity	Success/failure trend over 14 days	Positive trend	0.10

Output

Formatted scorecard with:

• Per-metric score, trend arrow, and status (HEALTHY / WARNING / CRITICAL)
• Composite health score (0–100)
• Ranked improvement opportunities with recommended action
• Weakest subsystem identification

Integration

• Memory: --store saves scorecard as episode ([[zouroboros.introspection]])
• Prescribe: Output feeds Phase 2 prescription engine
• Scheduling: Run as daily Zo agent at 05:00 AM Phoenix (after conversation capture at 04:00)

References

• references/metric-thresholds.md — threshold tuning guide and rationale